Hierarchical Digital Twin-Assisted RAN Resource Scheduling in Low-Altitude Scenarios

Radio Access Network (RAN) has been attracting wide attentions which can serve mobile applications to meet time-critical requirements in 6G scenarios, such as Uncrewed Aerial Vehicles (UAVs)-based parcel distributions. The reason behind this is to enable a RAN slicing technology to achieve customized resource scheduling for data transmission. Nonetheless, conventional RAN slicing solutions usually focus on RAN network optimization, while neglecting the impact of UAV data volume on slicing performance. To address this, we propose a Hierarchical Digital Twin (DT)-Assisted RAN Resource Scheduling (HDT-RAN) framework. In this framework, we derive two DT modules: UAV-DT and RAN-DT. The UAV-DT can imitate UAV state to dynamically adjust data compression ratios using a Contextual Multi-Armed Bandit (Contextual MAB) method. The RAN-DT can imitate RAN states to accurately predict network node loads with a hybrid-sampling attention mechanism. Experimental results show that our solution performs feasible compression ratios and reduces network loads significantly compared to baseline methods.